Section 2.6: How to Write a SAS Dataset

In order to apply SAS procedures or manipulate datasets, all data must 
first be placed into a SAS dataset. The DATA step is designed for this 
purpose. Usually, data will be entered into a spreadsheet such as Excel or 
into an ASCII text file (free or fixed format) to be read by SAS. If the 
number of records and/or variables is relatively "small", data can be 
placed directly within the SAS program following a CARDS; or DATALINES; 
statement at the bottom of the DATA step. Reading data into temporary 
datasets is an economical use of disk space and avoids the clutter of 
having many files.


Some important features of SAS datasets include:

1. SAS datasets can be either temporary or permanent. "Temporary" means 
they are stored in the 'work' directory and will disappear when you exit 
PC SAS for Windows (in the batch program terminates). "Permanent" means 
they have been saved in SAS dataset format in a file with a name on the 
hard disk (the suffix depends on the type of operating system and version 
of SAS which produced it).

2. The prefix consists of up to eight letters (a...z), numbers 
(0,1,...,9), or the underscore (_) which identifis a dataset name and 
briefly describes the contents of the file.

Note in PC SAS for windows if you write dataset names that begin with the 
underscore _, they will not be visible to the SAS EXPLORE window; however, 
they are placed in the designated directory. You can treat them the same 
as any other dataset, even though you cannot see their name. When the SAS 
system operates the underscore serves as the prefix to the temporary 
datasets which are intended to be hidden from SAS users. So don't be 
surprised if you can't see the datasets with this initial character that 
you may have generated in your program.

3. The first character of the dataset name must be a letter or the 
underscore, even when a permanent data set is created (e.g., a temporary 
data set called 1917s or a permanent data set <libref>.1917s are NOT valid 
dataset names - more about permanent datasets and the use of <libref> 
later).

Below are the keywords for the SAS DATA step statements that are needed to 
read an external space-delimited text file called mydata.dat on a PC 
system located in a subdirectory called c:\data\mydata:

DATA new;
INFILE "c:\data\mydata.dat";
INPUT name $ b c;
RUN;

In this short set of commands, a temporary SAS dataset called "new" is 
defined on the DATA statement. The external file to be read by SAS is 
called mydata.dat which is assumed to be in free text format (space 
delimited), located in the c:\data directory as defined on the INFILE 
statement. (On a unix system when running a program from your own user 
area the notation ~/data is a sufficient abbreviated version of the 
complete working directory.) The text file to be read in this example is 
called mydata.dat and contains three distinct variables called name 
(assumed to be 8 characters or less) and b c (both numeric) as listed on 
the INPUT statement in free format. In the datafile, they are to be 
separated by 1 or more spaces and have either the actual value or one of 
the 28 SAS missing data codes entered.

Temporary SAS datasets disappear after you exit Windows or run a batch 
program. With large or frequently accessed datasets, writing permanent SAS 
datasets (i.e., saved on the hard drive) may have several advantages. How 
to work with "permanent" SAS datasets will be described in Chapter 4.